Reasons for failur in automatic trading systems
- Thread starter cohvi
- Start date
??? Which part???
Does he write... edge sustaining regardless of market condition?
... Bull... Bear... in your example...
Anyways, your question can be answered by giving it some more thought into it.
You should test the same model against time period that is in bull market condition to see if the trading statistics are significantly different. If so, you know the setup is market condition dependent.
i.e. If the model is has an edge, then the sharpe ratio for both period should be similar.
Thank you all,
Until know I learned some interesting things from this thread but haven't really got a good answer to the main question asked above:
Why do most of the people fail in creating automatic trading machines?
There is the problem of backtesting vs future testing, and the approach of "Dynamic time analysis" suggested by Walther on other threads should supply the answer here.
Putting aside the ability to handle automatic actions and feed, the main thing is the method of trading. As I understand the advantage of exploiting edges in the market, I'm talking about taking successful simple profitable trading methods that people do use, methods that require only technical quantifiable parameters (High, Low, Open, Close, Volume) and automating them.
Now the main question is this:
Do we have trading strategies that can work for a long time (with maintenance)?
Not edge taking, but more like trend followers methods.
I think YES.
Until know I learned some interesting things from this thread but haven't really got a good answer to the main question asked above:
Why do most of the people fail in creating automatic trading machines?
There is the problem of backtesting vs future testing, and the approach of "Dynamic time analysis" suggested by Walther on other threads should supply the answer here.
Putting aside the ability to handle automatic actions and feed, the main thing is the method of trading. As I understand the advantage of exploiting edges in the market, I'm talking about taking successful simple profitable trading methods that people do use, methods that require only technical quantifiable parameters (High, Low, Open, Close, Volume) and automating them.
Now the main question is this:
Do we have trading strategies that can work for a long time (with maintenance)?
Not edge taking, but more like trend followers methods.
I think YES.
Cohvi
From my experiences I have not found a fully technical/auto/mechanical rule-based system to work on any single instrument but have found great success appying a strictly rule based system to a portfolio of equities which I select based on certain criteria. Once a stock fails to meet my requirements I remove it from my setup list and eventually add one that does.
1/They are just no good enough to design profitable system, even less than automate it.Profitable automating I saw in various outfits was basically ongoing fitting with quants going back and forth, adjusting like crazy.
2/Dynamic time analysis is a different approach to trading where one uses time only to identify profitable areas .
It is impossible to code and test.
3/You are correct.They exist . One can take basically any system or method and make it at least 50% more profitable when using dynamic time filters. What it does is it reduces number of trades , eliminating significant amount of losing trades.
Again , it is impossible to code. I saw decent size accounts with Sharpe > 2 using that stuff.
There are no completely automated trading systems that make money...how could there be? What "manual" system would you automate? If you could guarantee success with any system, then automate it, you would have all the money in the world, right?
Sure, we have developed and prospered by finding strategies that work (openings, scalpers, etc.) and automating a portion (or all aspects) for a short time...but the market figures it all out after a very short time.
Even Program trading at it's best has slippage and has to modify assumptions on a daily basis, right?
The best decision making computer is stuck right between your ears, and uses your eyes for entry / exit points.....quantitative strategies, such as pairs and mergers "work well" - but have to be babysat just like anything else.
I've had traders actually waste years of their lives, after making good money for years...basically sacrificing earnings trying to develop "systems" ....
And, as always..."be careful what you wish for" - for if it were possible, then there would be no viable marketplace any longer, right?
Trade em' - make money - modify and adapt to the marketplace.
Perhaps this may be of interest.
www.stocktrading.com/Adapting.html
All the best,
Don
And, before you guys get upset, please understand that I have many traders using automated programs....some work for a while, some never worked too well to begin with.
Sure, we have developed and prospered by finding strategies that work (openings, scalpers, etc.) and automating a portion (or all aspects) for a short time...but the market figures it all out after a very short time.
Even Program trading at it's best has slippage and has to modify assumptions on a daily basis, right?
The best decision making computer is stuck right between your ears, and uses your eyes for entry / exit points.....quantitative strategies, such as pairs and mergers "work well" - but have to be babysat just like anything else.
I've had traders actually waste years of their lives, after making good money for years...basically sacrificing earnings trying to develop "systems" ....
And, as always..."be careful what you wish for" - for if it were possible, then there would be no viable marketplace any longer, right?
Trade em' - make money - modify and adapt to the marketplace.
Perhaps this may be of interest.
www.stocktrading.com/Adapting.html
All the best,
Don
And, before you guys get upset, please understand that I have many traders using automated programs....some work for a while, some never worked too well to begin with.
You have made two assumptions above:
One, that the automated system will scale to support any size trades and two, that an automated trading system will continue to work forever.
Using the same flawed logic you could make the argument that there can't be any long-term successful traders otherwise they would eventually end up with all the money in the world. I'm sure you don't think that is true.